JP7494938B2 - Throughput estimation device, throughput estimation method, and program - Google Patents

Description

The present invention relates to a throughput estimation device, a throughput estimation method, and a program.

Various communication services (telephone, video distribution, Web, video conferencing, desktop virtualization, IoT, etc.) that transfer data such as video and audio (including voice) between terminals or between servers and terminals via a network have become widespread.

In communication services using video and audio, if network resource shortages, failures, or other problems occur, network quality (throughput, packet loss, packet transfer delays, etc.) will deteriorate, and the quality of video and audio experienced by the viewer (QoE: Quality of Experience) will also deteriorate.

Many video distribution services use adaptive bitrate video distribution, which changes the video and audio bitrates (amount of data per unit time when playing video and audio) according to the state of throughput (amount of data transferred per unit time). In this distribution method, video and audio data with different representations (a set of codec, bitrate, resolution, frame rate, etc. for video, a set of codec, bitrate, etc. for audio) are placed in advance on the distribution server. During distribution, the terminal requests a representation corresponding to an appropriate bitrate from the server each time depending on the throughput situation, and receives and plays video and audio while switching the bitrate. As a result, a decrease in throughput causes a decrease in image quality or sound quality due to the selection of a low bitrate, and the transfer of video and audio data required for playback cannot be kept up, resulting in a delay in playback start or playback stopping due to buffering processing caused by the depletion of video and audio data stored in the buffer of the receiving terminal, deteriorating QoE.

Deterioration in QoE affects viewer engagement (viewing time, viewing abandonment, viewing cancellation, etc.), but the QoE required to maintain appropriate engagement differs depending on various contexts, such as the user, content, and fee structure. Therefore, it is desirable to provide services with an appropriate QoE for each context.

Therefore, it is beneficial for video distribution providers to use a network that can provide sufficient throughput to meet the desired QoE (target QoE) in order to improve user engagement. It is also desirable for network operators to provide sufficient throughput to meet the target QoE in order to encourage video distribution providers to use their networks more frequently. However, if excessive throughput is provided, the target QoE can be met, but the network equipment costs will increase. Therefore, from the perspective of QoE and cost, it is important to understand the minimum required throughput and design and control the network based on that.

Therefore, there is a need for technology that can estimate the minimum required throughput to achieve a given QoE.

Conventionally, there is a technology shown in Patent Document 1 that models the relationship between throughput and QoE. This technology is related to a model that estimates QoE using throughput as input, and by utilizing the correspondence relationship between throughput and QoE in this model, it is possible to derive the throughput corresponding to any QoE.

Furthermore, there are technologies shown in Non-Patent Documents 1 and 2 that model the relationship between bitrate and QoE. This technology estimates QoE using quality parameters such as video and audio bitrates, video resolution, and video frame rate as input. Generally, a bitrate is selected that is equal to or smaller than the total value of the video and audio bitrates for the throughput, so the total value of the selected video and audio bitrates is regarded as the throughput, and by using a model of bitrate and QoE, it is possible to derive the corresponding throughput from the QoE.

JP 2019-121847 A

Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport, ITU-T P.1203 K. Yamagishi and T. Hayashi, "Parametric Quality-Estimation Model for Adaptive-Bitrate Streaming Services," IEEE Transactions on Multimedia, 2017. DOI: 10.1109/TMM.2017.2669859

However, the technology of Patent Document 1 estimates QoE using only throughput as input, and therefore does not take into account differences in representation. Adaptive bitrate video distribution changes the bitrate by switching the representation according to the throughput situation, but the selectable representations differ depending on the service, content, etc. Therefore, even if the throughput is the same, the same representation is not necessarily selected, and the QoE is not necessarily the same. For example, when the throughput is high and a representation including a high bitrate is selectable, the image quality is improved by selecting a high bitrate, and the QoE is high. On the other hand, even if the throughput is high, if a representation including a high bitrate is not available, the high bitrate is not selected, the image quality is not improved, and the QoE is not improved. Therefore, with existing technologies, it is difficult to accurately estimate the throughput depending on the representation.

In addition, in the technologies of Non-Patent Document 1 and Non-Patent Document 2, the relationship between QoE and throughput can be derived from a model of bit rate and QoE on the premise that a bit rate equivalent to the throughput is selected, but the capacity of a server is finite, and the number of selectable bit rates is limited to a few. Therefore, for a throughput section where a bit rate equivalent to the throughput does not exist, a bit rate far removed from the throughput is selected, or multiple bit rates are selected while being switched. Therefore, with the existing technologies that model the relationship between bit rate and QoE, it is difficult to accurately estimate throughput from QoE.

The present invention has been made in consideration of the above points and aims to improve the estimation accuracy of the throughput required to satisfy an arbitrary QoE.

Therefore, in order to solve the above problem, the throughput estimation device has a QoE estimation unit that estimates the QoE for each of a plurality of selection candidates based on each of the selection candidates for a parameter set related to the quality of video delivered over a network, and a throughput estimation unit that receives as input the QoE estimated by the QoE estimation unit for each selection candidate, the parameter set for each selection candidate, and a target QoE, and estimates the throughput required to satisfy the target QoE.

It is possible to improve the accuracy of estimating the throughput required to satisfy a given QoE.

1 is a diagram illustrating an example of a hardware configuration of a throughput estimation device 10 according to an embodiment of the present invention. 1 is a diagram illustrating an example of a functional configuration of a throughput estimation device 10 according to an embodiment of the present invention. FIG. 11 is a diagram showing a configuration example of representation information. 1 is a diagram showing an example of the configuration of QoE estimation information output from a QoE estimation unit 11. FIG. FIG. 1 is a diagram illustrating the relationship between QoE and throughput. 1 is a flowchart illustrating an example of a processing procedure executed by a throughput estimation device.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a diagram showing an example of a hardware configuration of a throughput estimation device 10 in an embodiment of the present invention. The throughput estimation device 10 in Fig. 1 has a drive device 100, an auxiliary storage device 102, a memory device 103, a CPU 104, an interface device 105, etc., which are interconnected by a bus B.

The program that realizes the processing in the throughput estimation device 10 is provided by a recording medium 101 such as a CD-ROM. When the recording medium 101 storing the program is set in the drive device 100, the program is installed from the recording medium 101 via the drive device 100 into the auxiliary storage device 102. However, the program does not necessarily have to be installed from the recording medium 101, but may be downloaded from another computer via a network. The auxiliary storage device 102 stores the installed program as well as necessary files, data, etc.

When an instruction to start the program is received, the memory device 103 reads out the program from the auxiliary storage device 102 and stores it. The CPU 104 executes functions related to the throughput estimation device 10 in accordance with the program stored in the memory device 103. The interface device 105 is used as an interface for connecting to a network.

Figure 2 is a diagram showing an example of the functional configuration of the throughput estimation device 10 in an embodiment of the present invention. In Figure 2, the throughput estimation device 10 has a QoE estimation unit 11, a throughput estimation unit 12, etc., in order to estimate the required throughput (amount of data transferred from the server to the terminal per unit time) to satisfy the quality experienced by the user (QoE: Quality of Experience) for adaptive bitrate video distribution. Each of these units is realized by a process in which one or more programs installed in the throughput estimation device 10 are executed by the CPU 104. In other words, each of these units is realized by cooperation between the hardware resources of the throughput estimation device 10 and the programs (software) installed in the throughput estimation device 10.

The QoE estimation unit 11 receives representation information as input, estimates the QoE of each representation based on the representation information, and outputs the representation information and the estimated QoE.

Figure 3 is a diagram showing an example of the configuration of representation information. The representation information includes one or more representations that can be selected (candidates for selection) for a certain video distribution service (hereinafter referred to as the "target service") for which the minimum required throughput is to be estimated. One representation is composed of a set of parameters (parameter set) related to the quality of the video or audio to be distributed, such as video bit rate, video resolution, video frame rate, audio bit rate, etc.

Video bit rate and audio bit rate are the set values for the amount of data per unit time for encoded video or audio data, respectively. Video resolution is the number of pixels per frame (number of vertical pixels x number of horizontal pixels). Video frame rate is the number of frames per second.

The representation information of the target service can be acquired by the server of the target service or the terminal that uses the target service. When the throughput estimation device 10 is installed in a location other than a server or terminal (e.g., on a network), the QoE estimation unit 11 acquires the representation information through communication with the server or terminal. Alternatively, the throughput estimation device 10 may be installed on a network, and the correspondence between network information such as a 5-tuple (source IP address, destination IP address, protocol, source port, destination port) and the representation information may be stored in advance in a DB or the like, and the representation information of the target service may be acquired by referencing the DB from the network information of the target service.

The QoE estimation unit 11 calculates an estimated QoE value (hereinafter simply referred to as "QoE") for each representation included in the acquired representation information using a QoE estimation model. As the QoE estimation model, an existing technology such as ITU-T Recommendation P. 1203 that estimates QoE using the video bit rate, video resolution, video frame rate, and audio bit rate as input may be used.

Figure 4 is a diagram showing an example of the configuration of QoE estimation information output from the QoE estimation unit 11. Figure 4 shows an example of QoE estimation information output by the QoE estimation unit 11 when Figure 3 is the input. As shown in Figure 4, the QoE estimation information includes, for each representation, the QoE calculated by the QoE estimation unit 11 for that representation.

The throughput estimation unit 12 receives the target QoE and the QoE estimation information output from the QoE estimation unit 11 as input, and estimates the minimum throughput required to satisfy the target QoE. The target QoE refers to the QoE that is targeted in order to improve user engagement in the target service.

Before explaining the method for estimating throughput, we will explain the relationship between QoE and throughput. Figure 5 shows the relationship between QoE and throughput.

The coordinate system shown in FIG. 5 is a coordinate system in which the horizontal axis corresponds to QoE, and the vertical axis corresponds to the throughput or bit rate corresponding to QoE. The dotted line shows the relationship between the bit rate and QoE when no playback stop occurs. In this case, it is known that as the bit rate becomes smaller or larger, an S-shaped curve converges to a certain QoE value. Therefore, in order to satisfy the target QoE, it is sufficient that a bit rate corresponding to the target QoE is selected and no playback stop occurs. Although it depends on the bit rate selection algorithm of adaptive bit rate video distribution, generally, if there is an equal or slightly higher throughput for any bit rate, that bit rate is selected. However, in reality, the bit rate options are limited. Also, the variation of selectable bit rates differs depending on the service and content. FIG. 5 shows an example in which there are three selectable bit rates. QoE _i and BR _i indicate the i-th QoE and the i-th bit rate in the representation list.

The plot points (black circles) in Figure 5 are uniquely determined based on the QoE estimation information output by the QoE estimation unit 11. The relationship between throughput and QoE between these plot points has the property (characteristic) of passing through both plot points and forming an upward convex curve, as shown by the solid line. Therefore, as shown in Figure 5, the relationship between bit rate and QoE (dotted line) and the relationship between throughput and QoE are different, so the throughput to satisfy the target QoE cannot be correctly derived only from the relationship between bit rate and QoE.

Therefore, in this embodiment, the throughput estimation unit 12 obtains plot points from the output result of the QoE estimation unit 11, and uses an estimation model that takes into account the above-mentioned properties to complement the intervals between the plot points (intervals between each QoE estimated by the QoE estimation unit 11), thereby estimating a throughput to satisfy the target QoE for all QoEs for various representation lists. Note that when the QoE is smaller than the minimum value of QoE _i (QoE value to the left of QoE ₁ in FIG. 5) or larger than the maximum value of QoE _i (QoE value to the right of QoE ₃ in FIG. 5), there is no throughput that satisfies the QoE in the given representation list without stopping playback, so in this embodiment, the target QoE is excluded from the input, or the value of the target QoE is changed (e.g., the target QoE is replaced with the QoE _i with the closest value).

The following is an estimation model formula for Throughput ₀ , which is the throughput between QoE _i and QoE _i+1 , assuming conditions where there is little fluctuation in throughput, such as in a bandwidth guaranteed network.

ここで、ＱｏＥ_{ｔａｒｇｅｔ}は目標ＱｏＥを示す。ａは係数を示す。Ｌｉｎｅの数式は、点（ＱｏＥ_ｉ＋１，ＢＲ_ｉ＋１）と点（ＱｏＥ_ｉ，ＢＲ_ｉ）を通る直線を表しており、Ｃｕｒｖｅの数式は、点（ＱｏＥ_ｉ＋１，０）と点（ＱｏＥ_ｉ，０）（すなわち、ＱｏＥ_ｉ＋１及びＱｏＥ_ｉの２つのＱｏＥに対応する横軸上の２つの点）を通る凸型の曲線を表しており、ＬｉｎｅとＣｕｒｖｅを加算することで、点（ＱｏＥ_ｉ＋１，ＢＲ_ｉ＋１）と点（ＱｏＥ_ｉ，ＢＲ_ｉ）を通る凸型の曲線が得られる。なお、当該曲線は、他の２点を通る凸型の曲線式で代替されても構わない。

Here, QoE _target indicates the target QoE. a indicates a coefficient. The formula of Line represents a straight line passing through points (QoE _i+1 , BR _i+1 ) and (QoE _i , BR _i ), and the formula of Curve represents a convex curve passing through points (QoE _i+1 , 0) and (QoE _i , 0) (i.e., two points on the horizontal axis corresponding to two QoEs, QoE _i+ 1 and QoE _i ), and by adding Line and Curve, a convex curve passing through points (QoE _i+1 , BR _i+1 ) and (QoE _i , BR _i ) is obtained. Note that this curve may be replaced with a convex curve formula passing through other two points.

Next, we show an estimation model formula for throughput when assuming conditions in which throughput fluctuates greatly, such as in a best-effort network service, and throughput drops significantly.

ここで、β、γは係数を示す。β及びγの値は、提供されるネットワークのスループットの安定性に応じて設定される。例えば、当該スループットの分散、標準偏差、信頼性区間などを参考にしてβ及びγの値を設定することができる。また、共有型のネットワークについて複数の通信を１つの回線に束ねてネットワーク設計及び制御を行う場合、統計多重化効果により、スループットの変動が他のスループットの変動に吸収されるため、β、γのそれぞれが０、１に近い値に設定されても構わない。更に、ネットワークの設計及び制御側でスループットの変動を加味して、スループットの下限を一定以上に保つ場合、β、γそれぞれの値が０、１に設定されてもよい。なお、Ｔｈｒｏｕｇｈｐｕｔは、Ｔｈｒｏｕｇｈｐｕｔ_０の上振れ分を考慮したものであり、他の線形式（二次関数、三次関数など）や非線形式（対数関数など）によってＴｈｒｏｕｇｈｐｕｔの推定モデル式が代替しても構わない。

Here, β and γ indicate coefficients. The values of β and γ are set according to the stability of the throughput of the provided network. For example, the values of β and γ can be set with reference to the variance, standard deviation, and reliability interval of the throughput. In addition, when a shared network is designed and controlled by bundling multiple communications into one line, the statistical multiplexing effect causes the fluctuations in throughput to be absorbed by the fluctuations in other throughputs, so β and γ may be set to values close to 0 and 1, respectively. Furthermore, when the network design and control side takes into account the fluctuations in throughput and keeps the lower limit of the throughput at a certain level or higher, the values of β and γ may be set to 0 and 1, respectively. Note that Throughput takes into account the upward swing of Throughput ₀ , and the estimation model formula of Throughput may be replaced by other linear expressions (quadratic functions, cubic functions, etc.) or nonlinear expressions (logarithmic functions, etc.).

The following describes the processing procedure executed by the throughput estimation device 10. Figure 6 is a flowchart for explaining an example of the processing procedure executed by the throughput estimation device 10.

In step S101, the QoE estimation unit 11 calculates the QoE for each representation included in the representation information of the target service (S101). The QoE estimation unit 11 generates QoE estimation information by assigning the calculated QoE to each representation included in the representation information, and inputs the QoE estimation information to the throughput estimation unit 12.

Next, the throughput estimation unit 12 calculates a throughput required to satisfy the target QoE given as input information based on the QoE estimation information (S102). That is, the throughput estimation unit 12 specifies each point in Fig. 5 based on the QoE estimation information, and calculates Throughput 0 or Throughput by substituting the target QoE, QoE _i (maximum QoE smaller than the target QoE) and QoE _i+1 (minimum QoE larger than the target QoE) of the section including the target QoE, BRi, and BRi+1 into any of the _above -mentioned estimation model formulas.

As described above, according to this embodiment, the target QoE and the representation information are used to estimate the throughput. Therefore, it is possible to improve the accuracy of estimating the throughput required to satisfy any QoE, such as the target QoE.

Therefore, according to this embodiment, it is possible to grasp the throughput required to satisfy the target QoE, and by designing and controlling the network based on that, it is possible to provide a network that satisfies the target QoE.

Although the embodiments of the present invention have been described in detail above, the present invention is not limited to such specific embodiments, and various modifications and variations are possible within the scope of the gist of the present invention as described in the claims.

10: Throughput estimation device 11: QoE estimation unit 12: Throughput estimation unit 100: Drive device 101: Recording medium 102: Auxiliary storage device 103: Memory device 104: CPU
105 Interface device B bus

Claims (6)

a QoE estimation unit that estimates a QoE for each of a plurality of selection candidates for a parameter set related to the quality of a video delivered via a network, based on each of the selection candidates;
a throughput estimation unit which receives as input the QoE estimated by the QoE estimation unit for each of the selection candidates, the parameter set for each of the selection candidates, and a target QoE, and estimates a throughput required to satisfy the target QoE;
A throughput estimation device comprising: The throughput estimation unit estimates a throughput for satisfying the target QoE by using a model equation of the QoE and the throughput required to satisfy the QoE, the model equation being derived from the QoE for each of the selection candidates and the parameter set for each of the selection candidates.
2. The throughput estimation device according to claim 1. The model formula complements the interval between each QoE estimated by the QoE estimation unit based on the property that the relationship between the QoE and the throughput required to satisfy the QoE is a convex curve in a coordinate system in which the vertical axis corresponds to the throughput and the horizontal axis corresponds to the QoE.
3. The throughput estimation device according to claim 2. The model formula complements the section by adding a convex curve passing through points on the horizontal axis corresponding to the two QoEs of the section to a straight line passing through the section.
4. The throughput estimation device according to claim 3. a QoE estimation step of estimating a QoE for each of a plurality of selection candidates for a parameter set related to the quality of a video delivered via a network, based on each of the selection candidates;
a throughput estimation procedure that receives as input the QoE estimated for each of the selection candidates by the QoE estimation procedure, the parameter set for each of the selection candidates, and a target QoE, and estimates a throughput required to satisfy the target QoE;
A throughput estimation method, comprising: A program for causing a computer to function as a throughput estimation device according to any one of claims 1 to 4.

Images